The latest news from tax, legal tech, and AI vendors points to a simple message for professional-services firms: a tool is not the same thing as a workflow. Whether the issue is weak chatbot measurement, adoption after go-live, or a company restructuring around growth, firm leaders should focus on how AI is measured, used, and sustained inside day-to-day operations.
AI value depends on measurement, not just deployment
A recent IRS watchdog report says the agency cannot confidently tell whether its chatbot and live chat applications are helping taxpayers because the performance data it collects is unreliable and it has not analyzed the chat applications. For firm owners, that is a useful warning: if you cannot measure whether an AI workflow is reducing handoffs, improving response quality, or routing work correctly, you do not really know whether it is working.
Custom AI systems need clear evaluation criteria from the start. That means defining what success looks like for intake, triage, drafting support, research assistance, or matter and engagement routing before the workflow goes live. Otherwise, teams may keep using the tool without knowing whether it is actually saving time or improving consistency.
Go-live is the beginning of adoption, not the finish line
A legal technology article argues that technical implementation and operational transformation are not the same thing. A system can be deployed well and still fail if the firm's habits and workflows do not change. The article also describes a familiar pattern: a project is declared complete, but active usage later falls and old workarounds return.
That is especially relevant for agentic workflows in law and accounting firms. If an AI assistant is meant to handle intake, summarize requests, or draft first-pass responses, leaders need to manage adoption after launch with training, accountability, and process ownership. The goal is not just to turn the system on, but to make the new workflow the easy default.
Restructuring at AI vendors shows the market is still changing
News that Darrow cut roles as part of a strategic restructure, while saying it has had consistent year-over-year growth and is rebalancing team capabilities, is another reminder that even well-known AI platforms are still evolving. For buyers, that means vendor diligence should go beyond feature lists and demos.
Professional-services firms should ask how a platform supports long-term use, what internal resources it requires, and whether the tool can be adapted to the firm's actual process. Custom AI succeeds when it fits the way the firm works, not when the firm reshapes itself around a generic product.
What this means for custom AI in law and accounting firms
The practical takeaway is to start with a narrow workflow that has a clear owner, a clear use case, and a clear measure of success. That might be client intake, document triage, issue spotting, or first-draft support. Build evaluation into the workflow itself so the team can see whether the AI is helping.
From there, treat adoption as an operational program. Review usage, capture where staff still rely on old steps, and adjust the workflow until it becomes the normal way work gets done. That approach is more durable than chasing isolated tools, and it is the best path to useful automation in a professional-services firm.
- Do not launch AI without a way to measure whether it is helping the firm.
- Treat go-live as the start of adoption work, not the end of the project.
- Use narrow, owned workflows before expanding to broader agentic automation.
- Vet vendors for operational fit and long-term support, not just product features.
Sources watched
- U.S. Judge Allows SEC-Musk Deal Despite 'Significant Misgivings' (CPA Practice Advisor AI)
- Darrow Cuts Roles as Part of Strategic Restructure (Artificial Lawyer)
- Why Adoption Starts Where Go-Live Ends (Artificial Lawyer)
- Are IRS Chatbots Really Helping Taxpayers? (CPA Practice Advisor AI)
